Register FAQ / Rules Forum Spy Search Today's Posts Mark Forums Read
Go Back   MacRumors Forums > Apple Hardware > Desktops > Mac mini

Reply
 
Thread Tools Search this Thread Display Modes
Old Jan 19, 2013, 04:06 AM   #1
magebarf
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Post Corsair Vengeance 1866MHz CL10 & Kingston HyperX PnP 1600MHz CL9 16GB + Mac Mini 2012

Update; This started as isolated test of the Corsair Vengeance RAM, but has now grown to also include the Kingston HyperX PnP RAM. Kingston info starts in post #13 (http://forums.macrumors.com/showpost...2&postcount=13)

Finally managed to get a hold of the Corsair Vengeance 16GB 1866MHz CL10 Kit (CMSX16GX3M2A1866C10) to try it out in the 2012 Mini.

Popped it in, booted into Apple Hardware Test and verified that my suspicions was right; the 2012 Mini runs the RAM at 1600MHz.

Ran the Apple Hardware Test fine, then Booted up Memtest86+ to verify the timings: 9-10-9-26 (First picture attached). Let it go quickly for 1 pass, worked out fine.

Booted into OS X and ran iBench, NovaBench, XBench and Geekbench in that order. I attached the pictures of the summaries from iBench, NovaBench and XBench and the RAM tests really does get a boost from them.

The Geekbench results is available for comparison here: http://browser.primatelabs.com/geekb...549130/1548933

What can be seen is that the RAM performance of the Vengeance 1866MHz kits are really good. Some of the tests (especially NovaBench) seem to give a better score for simply having more RAM. None of the tests however did allocate the amount of memory that it should be noticeable with just the size increase.

I also ran bandwidth, both 32 bit and 64 bit, to try to isolate out a smaller part of memory and do bandwidth testing, but everything there fell within statistical margins, and could hardly be registered by visually inspecting the graph given as output, so I believe that in the end that tool is more CPU bound than RAM bound.

Word of caution here; After doing these tests last night, I once again booted up Memtest86+, and let it run over night. When I woke up I found that it had completed 4 passes, detecting 3 errors... (Last attachment)

I might've caught a bad kit, I'm not sure. I'm going to go over and try to replace my kit at the retailer, with my photos as evidence of it not being quite healthy. I will also have to see if I will be getting a new Vengeance 1866MHz CL10 Kit, or if I simply should exchange it for one of the Kingston 1600MHz CL19 Kits that has been reported as good by other forum members (but which is almost $30 more expensive at my retailer).

The Kingston 1600MHz Kit's (both the "regular" and the Low-Voltage) is running at 9-9-9-27 timings compared to the 9-10-9-26 timings of the Vengeance 1866MHz (when run at 1600MHz), so its not quite clear which one is theoretically faster. It all boils down to exactly how the application you are running uses the memory.
Attached Thumbnails
Click image for larger version

Name:	latency.jpg
Views:	118
Size:	153.5 KB
ID:	391211   Click image for larger version

Name:	ibench.png
Views:	216
Size:	174.0 KB
ID:	391212   Click image for larger version

Name:	NovaBench.png
Views:	138
Size:	117.4 KB
ID:	391213  

Click image for larger version

Name:	xBench.png
Views:	119
Size:	259.2 KB
ID:	391214   Click image for larger version

Name:	testerror.jpg
Views:	82
Size:	193.3 KB
ID:	391215  

Last edited by magebarf; Jan 20, 2013 at 04:27 PM. Reason: Add Jump link to post about Kingston RAM
magebarf is offline   0 Reply With Quote
Old Jan 19, 2013, 09:15 AM   #2
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
And to follow up; Went to retailer, had the RAM kit exchanged to a identical kit.

Came home, ran Apple Hardware Test, all fine just like first kit.

Booted up Memtest86+ again; 2 errors detected on first pass, final result before I gave up was 27 Errors a few percent from 4 passes completed.

Either I've been unlucky, and got two duds, or this RAM kit really does not like running in the 2012 Mini.

Will see what to do about it tomorrow when I visit the retailer again, but the idea to switch to a Kingston HyperX PnP RAM running at 1600MHz with CL9 feels more and more tempting.

Edit; Added photo of Memtest86+ result.
Attached Thumbnails
Click image for larger version

Name:	second-faulty-vengeance.jpg
Views:	27
Size:	248.6 KB
ID:	391693  

Last edited by magebarf; Jan 21, 2013 at 03:03 PM. Reason: Put in final results of Memtest86+. Added photo of results.
magebarf is offline   0 Reply With Quote
Old Jan 19, 2013, 07:26 PM   #3
spartacvs
macrumors member
 
Join Date: Jan 2013
I'd stay away from Corsair Ram, too much problems with them.
__________________
2012 13" cMBP 16GB Ram, 256GB Crucial M4 SSD
2008 13" White Macbook 2GB Ram, 60GB HDD
2009 Mac Mini 8GB Ram, 128GB Sandisk SSD
iPad 3 16GB
spartacvs is offline   1 Reply With Quote
Old Jan 19, 2013, 09:13 PM   #4
dasx
macrumors 65816
 
dasx's Avatar
 
Join Date: Jun 2012
Location: Barcelona
You used to overclock all your PCs, right?

Nice thread. Will be following. I would so much like to see you test like 20 different memory kits and actually see what kind of difference does a good or bad memory make.

EDIT: OH, and it'd be nice to make those sticks run at 1866MHz, because there's gotta be a way.
__________________
Grey computer which does complicated stuff.
Cell phone that can talk to me. (Seriously).
"Fancy quote from dead guy"
dasx is offline   0 Reply With Quote
Old Jan 20, 2013, 04:27 AM   #5
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by dasx View Post
You used to overclock all your PCs, right?

Nice thread. Will be following. I would so much like to see you test like 20 different memory kits and actually see what kind of difference does a good or bad memory make.

EDIT: OH, and it'd be nice to make those sticks run at 1866MHz, because there's gotta be a way.
Hehe, one could think so, but actually no.

I've not overclocked any of my rigs... I did however put a lot of energy in to the research to get the right pieces of equipment that would give the best performance without going above the specs. So, factory-oveclocked Graphics card is as far as I've gone.

Closest thing is probably all of the dirty "auto-overclocking" that various motherboards tend to do. Also, a few of fan control and monitoring tools from e.g. Asus and Gigabyte, has had similar tendencies where they try to bump clocks for performance. I usually try to turn those features off whenever I'm aware of them.

In general; I'm not aiming for better performance with cheaper components by exchanging its lifetime. Then I rather save up a bit more to get better specced gear.

Regarding the possibility of it running at 1866MHz; I started out trying to do a bit of research of other peoples findings with various memories in the 2012 Mini (http://forums.macrumors.com/showthread.php?t=1516840). When it comes to 1866MHz I've only seen a single report about it.

It seems the 8GB Kinsgton Kit specced for 1866MHz worked fine at that speed in the 2011 Mini, but when the same RAM was put in the 2012 Mini, it stepped down to 1600MHz instead.

And here, we come also come to spartacvs thoughts:

Quote:
Originally Posted by spartacvs View Post
I'd stay away from Corsair Ram, too much problems with them.
Yes In general I do agree with this... The issue here being; I'm rather heavy on memory usage, using my Mini for both programming and music creation, thus I'm really in need of 16GB at the worst moments.

Also I was really curious if one could get any 1866MHz RAM to run in the 2012 Mini.

My experiences also say that you should always get a Kit which has gone through QC together, as they tend to be balanced so theres not too much variance between RAM sticks. This gets even more important the higher frequency you are running them on.

Bottom line; There as far as I have found is only a single kit of 16GB RAM specced for 1866MHz; The Corsair Vengeance.

Even worse: No one had tested it out, and lived to tell others.
Possibly even worse: The RAM as I said worked out very nicely, until I put it down a full night of Memtest86+, so if one wouldn't have tested it that thoroughly the issues would've gone unnoticed.

Only way I see one, without a new EFI update from Apple, could fix the issues would be by rewriting the SPD of the RAM to make the Mini use even worse timings and/or increase voltage, as this could stabilize it. This is a surefire way to void your Warranty though, and the only tools to do this I've found is Windows based.

It might be that these RAM kits are really just on the edge of the specs and that the numbers in the SPD works fine until you really push the RAM. If so, a manual bump of voltage would solve things, but this cannot be done in Apple's EFI in the same way as other manufacturers expose in their BIOS/EFI menus.

Quote:
Originally Posted by dasx View Post
EDIT: OH, and it'd be nice to make those sticks run at 1866MHz, because there's gotta be a way.
This is what I'm curious about as well. The only way as I see it today that this will happen is when Apple does an update to their EFI, at which time the negotiation/SPD reading might be updated so that the Mini selects the 1866MHz mode instead.

I'll have to see if I get another replacement, and if so if it comes out without any errors after crunching Memtest86+ over night (I think that would be my minimum time it should withstand a Memtest86+ run).

In this case I'll be happily awaiting any EFI updates further on to see if after applying one of them does the jump to 1866MHz instead.

More likely, the retailer will get tired of me switching RAMs over and over (I'm already a bit tired of this myself) and I'll move to one of the Kingston kits specced for 1600MHz instead... Will go there in half an hour, so will have to see what happens today.
magebarf is offline   0 Reply With Quote
Old Jan 20, 2013, 04:49 AM   #6
Giuly
macrumors 68040
 
Giuly's Avatar
 
Join Date: Oct 2009
Location: That depends whether you ask for timezone, state of mind or GPS coordinates.
According to the specifications, the memory controller only supports 1333/1600MHz.

Also, memtest86+ can give false results if you run both modules at once, test them seperately.

Last edited by Giuly; Jan 20, 2013 at 05:02 AM.
Giuly is offline   0 Reply With Quote
Old Jan 20, 2013, 05:03 AM   #7
philipma1957
macrumors 603
 
Join Date: Apr 2010
It is strange that the 2011 minis would run the 1866 and not the 2012. It needs to be the corsair or the kingston plug n play , but it did work.

IT speed up windows bootcamp graphics in measurable terms. Yet the 1866 is blocked for this machine.


Maybe there will be a change with the 10.8.3 update. Frankly I would just pick the kingston up for now just to be sure that there is no other issue with the machine.
philipma1957 is online now   0 Reply With Quote
Old Jan 20, 2013, 05:03 AM   #8
Alekto
macrumors member
 
Join Date: Nov 2012
Location: Tokyo
I've also tried the Corsair RAM. Wasn't stable at all for me.

Exchanged it for the Kingston HyperX CL9 RAM 16GB. Fast as hell.
Alekto is offline   0 Reply With Quote
Old Jan 20, 2013, 09:34 AM   #9
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by Giuly View Post
According to the specifications, the memory controller only supports 1333/1600MHz.

Also, memtest86+ can give false results if you run both modules at once, test them seperately.
Yes, you are right about the ARK just specing 1333 and 1600MHz RAM. But going on that one, the 2011 Minis specs (both Server and high end non-server) only gives 1066 & 1333MHz;

http://ark.intel.com/products/53463/...up-to-2_90-GHz
http://ark.intel.com/products/52231/...up-to-3_40-GHz

They do however run pretty much any 1600MHz RAM you throw at them, as well as some of the 1866MHz RAM people have tried with.

Seeing as Ivy Bridge is an evolution of the Sandy Bridge memory support shouldn't really be worse.

Looking at reviews, it is fully possible to find people that has managed to run the same set of RAM at 1866MHz using the same CPU in itself (http://www.legitreviews.com/article/1932/7/)
Dell also gives you the option of kitting their 3720QM Precision LapTops with 1866MHz RAM (the M4700 and M6700 models had this option previosly, now they just seem to use the 3740QM instead, which according to Intel's ARK takes same RAM according to the spec; look at this base configuration running 1866MHz Memory: http://www.dell.com/us/business/p/pr...recision-m4700)

However, it is fully possible that Apple has put restrictions in their EFI (which is what I believe) to restrict RAM speeds to 1600MHz, something they apparently did not have in the 2011 Mini.

As for "False positives" there is no such thing when it comes to RAM. If you get a single bit of data you read back wrong, depending on when and where this happens it could either be fully fine, or you could be looking at corrupted files or kernel panics.

Running the sticks of RAM alone in Memtest86+ is only a valid test if you plan to be running a single stick of RAM later on. If you are going to use both modules, you better test them together, as this is the only way to get an indication if they are working as they should.

It is fully possible that the sticks of memory themselves are working fine, but your system does not provide quite enough voltage when all of the RAM slots are filled. But, if this is the case you will be getting memory errors later on as well, so you better test things as you intend to run them.

Not doing this will only hide the errors from you, which in the end only will bite yourself in the ass.

----------

Quote:
Originally Posted by philipma1957 View Post
It is strange that the 2011 minis would run the 1866 and not the 2012. It needs to be the corsair or the kingston plug n play , but it did work.

IT speed up windows bootcamp graphics in measurable terms. Yet the 1866 is blocked for this machine.


Maybe there will be a change with the 10.8.3 update. Frankly I would just pick the kingston up for now just to be sure that there is no other issue with the machine.
Yep, strange indeed.

My theory is that for the 2011 Minis Apple did not have any kind of restriction on this in the EFI, which Apple may have put in place for the 2012 Minis however.

Since the last few generations of Intel processor it is the CPU itself that contains the memory controller for the system (compared to it being placed in the north bridge a few generations back), so barring correct designed motherboards it is only up to the CPU in combination with the system firmware whether it be BIOS or EFI to negotiate/restrict/configure what speeds the RAM is running at.

As I mentioned in the previous post, other people has successfully run RAM at 1866MHz with the same CPU that the 2.6GHz i7 uses, so the CPU in itself should not be the stopping factor here.

I can sadly tell you that OS X does not start loading until after the hardware has been initialized, so 10.8.3 will not "unleash" speeds of 1866MHz.

We still need to wait happily for a new EFI update to come, and hope that Apple modifies their RAM "restrictions" or negotiations at system initialization.

To ensure that the system in itself did not have any major issues I put it with Memtest86+ over night with the stock 4GB RAM again, and after 11 passes without any errors I decided to write of the Mini itself as the source of issues.



----------

Quote:
Originally Posted by Alekto View Post
I've also tried the Corsair RAM. Wasn't stable at all for me.

Exchanged it for the Kingston HyperX CL9 RAM 16GB. Fast as hell.
Did you also try the 1866MHz RAM, or did you go with their 1600MHz kit?

The retailer followed my reasoning, that its probably some incompatibility between the Mini and the Corsair Vengeance kit, so today I exchanged it for the same Kingston HyperX CL9 16GB Kit.

Will put it into the Mini soon, and then leave it for Memtest86+ over night.
magebarf is offline   0 Reply With Quote
Old Jan 20, 2013, 01:22 PM   #10
palebluedot
macrumors 6502a
 
Join Date: Jun 2008
I doubt Apple would willingly update the EFI on a consumer product to allow those of us with interest to use 1866. I wish it was the case, but the direction of Apple seems to speak otherwise.

It's weird that the 2011 can use faster BUS speed and has a better GPU. Really making me consider getting a 2011 instead of a 2012. Hrm.
palebluedot is offline   0 Reply With Quote
Old Jan 20, 2013, 02:00 PM   #11
53x12
macrumors 68000
 
53x12's Avatar
 
Join Date: Feb 2009
Quote:
Originally Posted by palebluedot View Post
I doubt Apple would willingly update the EFI on a consumer product to allow those of us with interest to use 1866. I wish it was the case, but the direction of Apple seems to speak otherwise.

It's weird that the 2011 can use faster BUS speed and has a better GPU. Really making me consider getting a 2011 instead of a 2012. Hrm.
Well what do you want/need in a computer anyways? You could get something to hold you over until the Haswell Mini as that is the one worth waiting for.
53x12 is offline   0 Reply With Quote
Old Jan 20, 2013, 02:17 PM   #12
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by palebluedot View Post
I doubt Apple would willingly update the EFI on a consumer product to allow those of us with interest to use 1866. I wish it was the case, but the direction of Apple seems to speak otherwise.

It's weird that the 2011 can use faster BUS speed and has a better GPU. Really making me consider getting a 2011 instead of a 2012. Hrm.
Actually, this happens a lot of the time, small improvements put into the EFI that's not explicitly stated in the documentation once the "big issues" gets a pushed fix.

On my MacBook White early 2009, they upped the max limit on RAM from 4GB to 8GB with the last EFI release they made. That one even kept the same version number, just a small bump in build number, so if one wasn't lucky to wait with updating until the last one, you'd have to know what you were looking for.

The same computer has also been running DDR2 at 800MHz since I did my upgrade, even though their own documentation states it only being able to run 667MHz RAM; so when fully EFI updated, the computer officially only supporting 4GB @ 667MHz is actually running 8GB @ 800MHz.

What Apple say and what Apple do is not always the same.

With regards to the 2011/2012 comparison, just do a quick comparison on the Geekbench results (http://browser.primatelabs.com/mac-benchmarks). Even though being able to run memory at a higher frequency, the improvements in CPU puts the 2012 far ahead in general performance.

The only case where you'd wanna go for the 2011 is when you pretty much only run applications using OpenCL... This is the only time the increase in CPU performance doesn't outrun the discreet GPU.

Just compare the pure numbers, and you'll see that performance wise there's not a great leap, but as mentioned there's a bit of feature difference (main one being OpenCL):
http://www.videocardbenchmark.net/gp...D+P4000&id=963
http://www.videocardbenchmark.net/gp...D+6630M&id=277
magebarf is offline   0 Reply With Quote
Old Jan 20, 2013, 04:24 PM   #13
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Post Kingston HyperX PnP 16GB (2x8GB) 1600MHz CL9

Allright...

As mentioned, the retailer agreed with my thoughts, so instead of trying for a third set of Corsair Vengeance 1866MHz RAM, I went with the Kingston HyperX PnP 16GB (2x8GB) 1600MHz CL9 RAM (KHX16S9P1K2/16).

Ran the same procedure as last time; Installed the RAM and ran the quick Apple Hardware Test to see that it seemed okay.

Booted into Memtest86+ to get the timings, which in line with philipma1957's comments about the HyperX RAM indeed was running at 9-9-9-27. This time I also let Memtest86+ run for two full passes, to verify that it wasn't catching any issues already, came out without any errors detected (first picture).

After two passes, I booted into OS X, and ran the same tests; iBench, NovaBench, XBench and Geekbench.

So, now I had a set full set of benchmarks from three sets of RAM to compare;

Stock Apple 4GB: 11-11-11-28 @1600MHz.
Corsair Vengeance 16GB: 9-10-9-26 @1600MHz.
Kingston HyperX 16GB: 9-9-9-27 @1600MHz.

The second, third and fourth picture attached is the results from iBench, NovaBench and XBench side to side accordingly.

For the Geekbench comparisons;
Corsair Vengeance vs Stock 4GB: http://browser.primatelabs.com/geekb...549130/1548933
Kingston HyperX vs Stock 4GB: http://browser.primatelabs.com/geekb...557287/1548933
Kingston HyperX vs Corsair Vengeance:
http://browser.primatelabs.com/geekb...557287/1549130

It's back and forth which one has the upper hand. One cycle less of tRCD for one cycle more of tRAS does show a bit in the stream tests where I'm guessing the specific tests have a bit different memory access/write behaviors, thus there is a bit larger difference between the two RAMs in these tests than the others.

All together, not really that big of difference between the Corsair and Kingston, both of them are ahead of the stock 4GB RAM the Mini was delivered with, which I'm guessing is mainly due to the better timings.

After performing the tests, I put the Mini back into Memtest86+ and let it run over night again. In the morning, it had complete three passes without detecting any errors.

So... Performance wise; with the current EFI where the Corsair Vengeance is restricted to running at 1600MHz frequency (with very unclear future for ever being run at 1866MHz in the 2012 Mini) is very similar to the Kingston HyperX.

Stability wise, out of my small test batch of two kits, which both as far as I could tell by the packaging had been shipped to Sweden all the way from Australia (or do all Corsair packages come with Australian warranty information?) was quickly put down by the Kingstons.

If I'd known I was going to end up with the Kingstons from the start, or if the retailer would have had it in stock, I would probably have gone for the Low Voltage variant of the HyperX though, running at 1.35V which may decrease the heat generated. But that test I will leave for someone else to do.
Attached Thumbnails
Click image for larger version

Name:	kingston.jpg
Views:	24
Size:	157.0 KB
ID:	391699   Click image for larger version

Name:	iBench-triple.png
Views:	89
Size:	171.0 KB
ID:	391700   Click image for larger version

Name:	NovaBench-triple.png
Views:	66
Size:	108.2 KB
ID:	391701  

Click image for larger version

Name:	XBench-triple.png
Views:	52
Size:	347.2 KB
ID:	391702  

Last edited by magebarf; Jan 21, 2013 at 03:23 PM. Reason: Adding proper report of the Kingston RAM
magebarf is offline   0 Reply With Quote
Old Jan 21, 2013, 12:20 AM   #14
palebluedot
macrumors 6502a
 
Join Date: Jun 2008
Quote:
Originally Posted by 53x12 View Post
Well what do you want/need in a computer anyways? You could get something to hold you over until the Haswell Mini as that is the one worth waiting for.
When is Haswell Mini supposed to come out. Seeing as I have a custom gaming PC with 2xSSD and about 18x the power of the Mac Mini, I think I am set on computing power That said, I want a Mini for work (dual monitors) so I'd love to bite the bullet now. Just want to get the most spec bang for my buck.

The buying guide says the Mini update is usually 300 days and we are only like 75 days in?

----------

Quote:
Originally Posted by magebarf View Post
Actually, this happens a lot of the time, small improvements put into the EFI that's not explicitly stated in the documentation once the "big issues" gets a pushed fix.

On my MacBook White early 2009, they upped the max limit on RAM from 4GB to 8GB with the last EFI release they made. That one even kept the same version number, just a small bump in build number, so if one wasn't lucky to wait with updating until the last one, you'd have to know what you were looking for.

The same computer has also been running DDR2 at 800MHz since I did my upgrade, even though their own documentation states it only being able to run 667MHz RAM; so when fully EFI updated, the computer officially only supporting 4GB @ 667MHz is actually running 8GB @ 800MHz.

What Apple say and what Apple do is not always the same.

With regards to the 2011/2012 comparison, just do a quick comparison on the Geekbench results (http://browser.primatelabs.com/mac-benchmarks). Even though being able to run memory at a higher frequency, the improvements in CPU puts the 2012 far ahead in general performance.

The only case where you'd wanna go for the 2011 is when you pretty much only run applications using OpenCL... This is the only time the increase in CPU performance doesn't outrun the discreet GPU.

Just compare the pure numbers, and you'll see that performance wise there's not a great leap, but as mentioned there's a bit of feature difference (main one being OpenCL):
http://www.videocardbenchmark.net/gp...D+P4000&id=963
http://www.videocardbenchmark.net/gp...D+6630M&id=277
Good points and I hope you are right, but it leads one to wonder why they restrict it in the first place if the hardware is capable of running that BUS speed. Seems like it's a force handicap for no (technical) reason.
palebluedot is offline   0 Reply With Quote
Old Jan 21, 2013, 02:49 AM   #15
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by palebluedot View Post
Good points and I hope you are right, but it leads one to wonder why they restrict it in the first place if the hardware is capable of running that BUS speed. Seems like it's a force handicap for no (technical) reason.
The only reasons I could think of that makes sense is either Quality Control (stability tests), or that it requires additional work to enable the higher frequencies. None of them does really add up with how they did with the 2011 though.

It could be that when testing the 2011 Mini (specced for 1333MHz) they had access to enough of 1600MHz RAM that they could test with, so they "let it go" as it seemed to work. I also don't know if 1866MHz RAM was even on the horizon by then? When the 2012 model (specced for 1600MHz) was being tested it might've been that there were so few 1866MHz RAM modules, or that the ones they had did not work properly, that they decided to disable them.

The other option is that our assumption of it being restricted is the other way around; that the CPU itself only keeps to the specified frequencies, unless the init routines explicitly states that higher frequencies should be allowed. If so, it makes perfect sense that the same CPU running on other motherboards/in other systems may support it as their system firmware would definitely be different, and could have had this enabled.
magebarf is offline   0 Reply With Quote
Old Jan 21, 2013, 06:39 AM   #16
philipma1957
macrumors 603
 
Join Date: Apr 2010
Quote:
Originally Posted by magebarf View Post
The only reasons I could think of that makes sense is either Quality Control (stability tests), or that it requires additional work to enable the higher frequencies. None of them does really add up with how they did with the 2011 though.

It could be that when testing the 2011 Mini (specced for 1333MHz) they had access to enough of 1600MHz RAM that they could test with, so they "let it go" as it seemed to work. I also don't know if 1866MHz RAM was even on the horizon by then? When the 2012 model (specced for 1600MHz) was being tested it might've been that there were so few 1866MHz RAM modules, or that the ones they had did not work properly, that they decided to disable them.

The other option is that our assumption of it being restricted is the other way around; that the CPU itself only keeps to the specified frequencies, unless the init routines explicitly states that higher frequencies should be allowed. If so, it makes perfect sense that the same CPU running on other motherboards/in other systems may support it as their system firmware would definitely be different, and could have had this enabled.
Quote:
Originally Posted by palebluedot View Post
When is Haswell Mini supposed to come out. Seeing as I have a custom gaming PC with 2xSSD and about 18x the power of the Mac Mini, I think I am set on computing power That said, I want a Mini for work (dual monitors) so I'd love to bite the bullet now. Just want to get the most spec bang for my buck.

The buying guide says the Mini update is usually 300 days and we are only like 75 days in?

----------



Good points and I hope you are right, but it leads one to wonder why they restrict it in the first place if the hardware is capable of running that BUS speed. Seems like it's a force handicap for no (technical) reason.


Not sure as to the why. There was not a lot of laptop 1866 plug n play ram for the 2011 mini or the 2012 mini. This is one of the times I would like to have the ear of an Apple engineer just for the sake of satisfying my curiosity.
philipma1957 is online now   0 Reply With Quote
Old Jan 21, 2013, 03:23 PM   #17
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Just put in a "proper" report of the Kingstons in post #13...
magebarf is offline   0 Reply With Quote
Old Jan 21, 2013, 08:24 PM   #18
paulrbeers
macrumors 68040
 
Join Date: Dec 2009
I'd love to see someone do a benchmark running 1333mhz RAM. Just curious is all. Looks like the decrease in latency from the Stock RAM to the Kingston is marginal. Of course we also don't know how much of the increase in speed is solely due to latency since the benchmark was done with two different amounts of RAM (who knows if the 4GB test was hitting the max RAM during the tests). Either way, doesn't really justify the extra $50 (difference right now between "normal" 1600mhz and the Kingston HyperX) just to get less than 3% faster benchmarks (geekbench anyway). Even the GPU benchmarks from Xbench only shows a minor increase and some of that could have been the fact that OSX allocates more system RAM to the GPU when more than 4GB is put in.

EDIT: Before anyone questions me on my thought process, 8-10 years ago, I used to buy expensive RAM trying to get every small percentage point out of my system. Ya know what I realized? In the end getting a 3% faster system at 6-8% of the overall system cost was silly. A buddy and I did a blind real world test (sat down played some FPS, surfed the internet, ripped some movies, etc.) and in the end I couldn't tell if I was running the fast RAM or the "slow" because you just don't notice 3-5% difference. 1 FPS is near impossible to tell, a video encoding finishing 30 seconds faster after it had run for an hour really wasn't noticeable, and in the end a Video plays back just the same whether on the faster RAM or the slower. Just my .02 worth.
__________________
rMBP 11,2 15" 2.0Ghz w/ 256GB SSD
MBA 6,2 13" 1.7Ghz w/ 256GB SSD
Mac Mini 6,2 2.3ghz w/ 240GB SSD + 1TB
Mac Pro 1,1 w/ 8 cores @ 2.66 w/ 240GB SSD

Last edited by paulrbeers; Jan 21, 2013 at 08:29 PM.
paulrbeers is offline   0 Reply With Quote
Old Jan 22, 2013, 03:50 AM   #19
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by paulrbeers View Post
I'd love to see someone do a benchmark running 1333mhz RAM. Just curious is all. Looks like the decrease in latency from the Stock RAM to the Kingston is marginal. Of course we also don't know how much of the increase in speed is solely due to latency since the benchmark was done with two different amounts of RAM (who knows if the 4GB test was hitting the max RAM during the tests). Either way, doesn't really justify the extra $50 (difference right now between "normal" 1600mhz and the Kingston HyperX) just to get less than 3% faster benchmarks (geekbench anyway). Even the GPU benchmarks from Xbench only shows a minor increase and some of that could have been the fact that OSX allocates more system RAM to the GPU when more than 4GB is put in.
Well, the actual latency numbers can easily be calculated, as we do know the RAM timings. What I don't know however is if the Apple 16GB RAM has the same timings as the 4GB set.

But, the stock 4GB having 11-11-11-28 timings, vs the 9-9-9-27 of the Kingston, it is just to calculate backwards from the frequency (as the numbers are cycles).

A bit harder is to evaluate what it actually turns out to in practice.

As for differences due to 4/16GB RAM; I kept processes to a minimum when benching, so I can't say if any of them pushed above free RAM from the 4GB kit. I will run them later tonight with the Activity Monitor up, to see exact amount of RAM usage, to verify that this wasn't the case (which i do not believe). As I've mentioned though, I got the feeling that at least NovaBench gave arbitrary extra points for just having more RAM, even though it's actual performance test did not give all that much difference.

As for the GPU tests though, none of them is really pushing the RAM bandwidth as far as I can tell, but to me seems more to be testing actual calculation performance. Seeing as XBench is a bit over 6 years old already, its also a bit outdated in the graphics tests I'm guessing.

In the optimal outcome, where the Corsair RAM would've actually run in 1866MHz (which would be a even better outcome than I was expecting) I believe that the price difference would've been worth it for some people. For me, I felt that the documentation of it was worth a bit extra seeing as I could not find anyone that had actually tested the RAM.

The jump from the Corsair Vengeance to the Kingston HyperX was more of an "best way out", trying to get fully working RAM. Seeing as it was the only proper 16GB kit (or Apple "certified" 8GB RAM sticks) my retailer had in stock that I could find any success stories for it was my easy way out.

If I would've contrasted e.g. the Crucial 16GB 1600MHz CL11 Kit (which is half the price) against the Kingston HyperX from the beginning, I'm much less likely to have gone for the more expensive one. But compared to the 1866MHz Vengeance with all of its unknown variables when paired with the 2012 Mini, the choice was not that straight forward.

As I said in the other thread; I've taken one for the team, and now with these experiences documented it will hopefully be easier for other people to decide if they think the same experimentation is worth it.

Quote:
Originally Posted by paulrbeers View Post
EDIT: Before anyone questions me on my thought process, 8-10 years ago, I used to buy expensive RAM trying to get every small percentage point out of my system. Ya know what I realized? In the end getting a 3% faster system at 6-8% of the overall system cost was silly. A buddy and I did a blind real world test (sat down played some FPS, surfed the internet, ripped some movies, etc.) and in the end I couldn't tell if I was running the fast RAM or the "slow" because you just don't notice 3-5% difference. 1 FPS is near impossible to tell, a video encoding finishing 30 seconds faster after it had run for an hour really wasn't noticeable, and in the end a Video plays back just the same whether on the faster RAM or the slower. Just my .02 worth.
I am fully with you on your reasoning, and this is most usually what the sane parts of my head tell me as well... But then sometimes the curious parts kick in and tell me otherwise.

In this case, curiosity got the upper hand. Most probably due to mentioned lack of documented experiences from others.

Last edited by magebarf; Jan 22, 2013 at 04:00 AM.
magebarf is offline   0 Reply With Quote
Old Jan 22, 2013, 07:19 AM   #20
paulrbeers
macrumors 68040
 
Join Date: Dec 2009
Quote:
Originally Posted by magebarf View Post
As I said in the other thread; I've taken one for the team, and now with these experiences documented it will hopefully be easier for other people to decide if they think the same experimentation is worth it.
.
Well, on behalf of the forums, we appreciate you taking one for the team! I finally bit the bullet and ordered a 2.3ghz Mid-Mini. I was going to hold out for Haswell (for no other reason than I didn't need a new mini at this time), but my Mac Pro isn't handling everything I am throwing at it anymore (seems to crash about once a week), so I decided it was time. I'm actually thinking about buying a set of the 1333mhz 16GB sets from newegg just to see how they affect performance compared to 1600mhz RAM. If it affects it too much, I'll throw them into my wife's 2011 Macbook Pro and buy the 1600mhz.

Edit: Nevermind I guess! The 1333mhz RAM I was going to buy from Newegg for $55 is no longer in their inventory.
__________________
rMBP 11,2 15" 2.0Ghz w/ 256GB SSD
MBA 6,2 13" 1.7Ghz w/ 256GB SSD
Mac Mini 6,2 2.3ghz w/ 240GB SSD + 1TB
Mac Pro 1,1 w/ 8 cores @ 2.66 w/ 240GB SSD
paulrbeers is offline   0 Reply With Quote
Old Jan 22, 2013, 07:27 AM   #21
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by paulrbeers View Post
Well, on behalf of the forums, we appreciate you taking one for the team! I finally bit the bullet and ordered a 2.3ghz Mid-Mini. I was going to hold out for Haswell (for no other reason than I didn't need a new mini at this time), but my Mac Pro isn't handling everything I am throwing at it anymore (seems to crash about once a week), so I decided it was time. I'm actually thinking about buying a set of the 1333mhz 16GB sets from newegg just to see how they affect performance compared to 1600mhz RAM. If it affects it too much, I'll throw them into my wife's 2011 Macbook Pro and buy the 1600mhz.

Edit: Nevermind I guess! The 1333mhz RAM I was going to buy from Newegg for $55 is no longer in their inventory.
Aww, too bad. Could've been an interesting analysis to do as well.
magebarf is offline   0 Reply With Quote
Old Jan 22, 2013, 07:31 AM   #22
paulrbeers
macrumors 68040
 
Join Date: Dec 2009
Quote:
Originally Posted by magebarf View Post
Aww, too bad. Could've been an interesting analysis to do as well.
Well, I just found 1333mhz 16GB set for $48 on Amazon and ordered it. As soon as it and my Mini arrive, I'll run some benchmarks!
__________________
rMBP 11,2 15" 2.0Ghz w/ 256GB SSD
MBA 6,2 13" 1.7Ghz w/ 256GB SSD
Mac Mini 6,2 2.3ghz w/ 240GB SSD + 1TB
Mac Pro 1,1 w/ 8 cores @ 2.66 w/ 240GB SSD
paulrbeers is offline   0 Reply With Quote
Old Jan 22, 2013, 07:54 AM   #23
magebarf
Thread Starter
macrumors member
 
Join Date: Dec 2012
Location: Stockholm - Sweden
Quote:
Originally Posted by paulrbeers View Post
Well, I just found 1333mhz 16GB set for $48 on Amazon and ordered it. As soon as it and my Mini arrive, I'll run some benchmarks!
Your starting point will be the same as mine, right? 4GB Stock RAM?
If so then we'll really be able to tell if any of the improvements is purely based on amount of RAM alone or if it is because of the timings.
magebarf is offline   0 Reply With Quote
Old Jan 22, 2013, 07:59 AM   #24
paulrbeers
macrumors 68040
 
Join Date: Dec 2009
Quote:
Originally Posted by magebarf View Post
Your starting point will be the same as mine, right? 4GB Stock RAM?
If so then we'll really be able to tell if any of the improvements is purely based on amount of RAM alone or if it is because of the timings.
Yes. I won't have a set of 16GB 1600mhz sticks. However if the 1333mhz 16GB set ends up showing the system to be slower then we know the "lack of RAM" has no affect on performance, but if the benchmarks show that the 1333mhz 16GB has equal or better performance then we know that the lack of RAM (4GB) is a bottleneck. Probably be a week before I get everything.
__________________
rMBP 11,2 15" 2.0Ghz w/ 256GB SSD
MBA 6,2 13" 1.7Ghz w/ 256GB SSD
Mac Mini 6,2 2.3ghz w/ 240GB SSD + 1TB
Mac Pro 1,1 w/ 8 cores @ 2.66 w/ 240GB SSD
paulrbeers is offline   0 Reply With Quote

Reply
MacRumors Forums > Apple Hardware > Desktops > Mac mini

Tags
1866mhz, corsair, mac mini 2012, ram, vengance

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Similar Threads
thread Thread Starter Forum Replies Last Post
Newegg Corsair Vengeance 16GB for 89.99 Until 3:00 PM EST paulrbeers Mac mini 22 Mar 31, 2013 07:13 AM
Anyone has the Corsair 16GB Vengeance Performance So-Dimm 1600MH? MaxPower72 MacBook Pro 3 Sep 23, 2012 04:46 PM
Corsair Vengeance 16GB (2x8GB) ScottishSkater MacBook Pro 16 Jun 6, 2012 09:46 PM

Forum Jump

All times are GMT -5. The time now is 09:23 AM.

Mac Rumors | Mac | iPhone | iPhone Game Reviews | iPhone Apps

Mobile Version | Fixed | Fluid | Fluid HD
Copyright 2002-2013, MacRumors.com, LLC